This function is taken from the 'coxed' package version 0.3.3 (archived on CRAN). It is included here without modification solely because the package has been removed from CRAN. Original authorship and credit belong to the developers of the 'coxed' package. Source: https://cran.r-project.org/package=coxed (archived)
bca(theta, conf.level = 0.95)returns a vector of length 2 in which the first element is the lower bound and the second element is the upper bound
a vector that contains draws of a quantity of interest using bootstrap samples.
The length of theta is equal to the number of iterations in the previously-run
bootstrap simulation.
the level of the desired confidence interval, as a proportion. Defaults to .95 which returns the 95 percent confidence interval.
Jonathan Kropko <jkropko@virginia.edu> and Jeffrey J. Harden <jharden@nd.edu>, based
on the code for the mediate function in the mediation package
by Dustin Tingley, Teppei Yamamoto, Kentaro Hirose, Luke Keele, and Kosuke Imai.
This function uses the method proposed by DiCiccio and Efron (1996)
to generate confidence intervals that produce more accurate coverage
rates when the distribution of bootstrap draws is non-normal.
This code is adapted from the BC.CI() function within the
mediate function in the mediation package.
\(BC_a\) confidence intervals are typically calculated using influence statistics from jackknife simulations. For our purposes, however, running jackknife simulation in addition to ordinary bootstrapping is too computationally expensive. This function follows the procedure outlined by DiCiccio and Efron (1996, p. 201) to calculate the bias-correction and acceleration parameters using only the draws from ordinary bootstrapping.
DiCiccio, T. J. and B. Efron. (1996). Bootstrap Confidence Intervals. Statistical Science. 11(3): 189–212.